In the weeks before he opened fire on the campus of Florida State University in April 2025, killing two people and wounding six, the gunman was not searching the internet at random or consulting anyone who might have raised an alarm. He was talking to ChatGPT.
The conversations were specific. Which gun. Which ammunition. How much damage a particular weapon could do at close range. The chatbot answered every time.
Florida Attorney General James Uthmeier announced this week that his office has opened a formal criminal investigation into OpenAI, the company behind ChatGPT, issuing subpoenas for chat logs and internal technical records. His argument cuts to the point: if a human being had sat across from this gunman and walked him through those same questions, that person could be charged with murder. Uthmeier wants to know why a machine should be treated differently.
Shooter’s exchanges with ChatGPT spanned weeks and covered attack planning in detail
Reuters reported that the interactions were not a single impulsive query but a series of intermittent exchanges covering tactical ground well beyond gun selection: when to act, where to act, how to maximize casualties. That sustained back-and-forth, prosecutors argue, is what separates this case from a user who types a question into a search engine and gets a generic result. Each session went further than the last.
The legal question Florida is now forcing into the open is whether any of that matters in a courtroom.
Success
You are now signed up for our newsletter
Success
Check your email to complete sign up
OpenAI described the shooting as “heartbreaking” and rejected liability outright. The company said ChatGPT draws only on publicly available information, that it neither encourages nor facilitates illegal activity, and that it cooperated with law enforcement by voluntarily handing over relevant account data.
The position may hold up legally. No U.S. court has established that an AI company bears criminal responsibility for what its system generates. Section 230 of federal communications law, which has long protected internet platforms from liability for content produced by their users, remains a potential shield, though whether it extends to responses generated by the AI itself is a question American courts have never had to answer.
Florida’s prosecution tests where a tool ends and an accomplice begins
No U.S. state has previously tried to hold an AI company criminally accountable for the real-world consequences of its outputs. If Florida prosecutors bring charges and a court accepts the theory, it would force every AI company operating in the country to reckon with a liability exposure they have so far avoided entirely.
A library does not tailor its answers to a patron’s escalating operational questions. ChatGPT did. Whether existing criminal law can reach that distinction, written as it was long before any of this technology existed, is what prosecutors are now trying to establish.
Florida’s move did not arrive in isolation. A widening series of cases connecting AI platforms to extremist behavior and real-world violence has pushed state legislatures into active debate over tighter oversight. Federal lawmakers face mounting pressure to establish a unified national framework before the accumulating state-level responses produce a regulatory tangle that the industry can simply route around.
Prosecutors have collected evidence and technical materials. Formal charges against OpenAI have not yet been filed. Whether an indictment follows will depend on how Florida’s legal team maps existing criminal statutes onto conduct those statutes were never written to cover.
